01 July 2003
Mesh meets the need
Wireless for industrial control.
By Adrian Tuck and Charlotte Burgess
It should be obvious that in an ever-growing number of industrial sensing and control applications, the switch from wired to wireless data communications is coming. Wireless reduces installation costs, adds flexibility, and eases deployment and maintenance hassles. Industrial publications and conference proceedings are full of articles prophesying the rise of wireless, but the reality is that most industrial customers have been reluctant to make the change. Why?
Benefits of wireless
Talk to anyone who has tried to make wireless work in an industrial environment, and you will hear about two big interrelated problems, usually in the form of a war story that ends in lost funds and lost time. Although wireless systems may seem like an obvious solution for industrial applications, in reality the cure can be worse than the disease.
Reliability means different things to different people, and this is the root of the problem for industrial customers. When confronted with a patchy signal on a cell phone, a user moves around until the signal clears. Translate this situation to an industrial environment, where the wireless device is a temperature sensor stuck onto a pipe, and you can see how standard wireless technology fails. Are you going to move the pipe closer to a window?
Most wireless networks do not offer the day-to-day, month-to-month, year-to-year type of reliability that you get from a wired system because most wireless systems use a hub-spoke (star) topology where devices talk to a base station (802.11 and Bluetooth both use this topology). Data has only one possible path to travel in this model and depends on both the constant proper operation of the base station and the elimination of any interference between the wireless sensor and the base station. Unfortunately, especially in industrial environments, it is very difficult to mandate the quality of the air between two devices. The changing noise floor of radio interference leads to the next big problem.
Most existing wireless technology re-quires considerable deployment time to try to mitigate the risks of interference. If there is only one path from the temperature sensor to the base station, then you need to fix the position of each to a point that minimizes the risk of downtime. This translates to significant investments in radio frequency (RF) site surveys, usually requiring specialist knowledge. And here's the rub—just because it works on day one, that doesn't mean it is going to work on day two. Equipment moves, interference patterns change, and as a result, networks can fail.
Of these problems, poor reliability is the real showstopper. But why is reliability so hard to achieve? Most of us have experienced the symptoms of it with cell phones:
- Signals appear and disappear simply because you move your phone six inches.
- You can hear the other person, but they can't hear you.
- Calls get dropped and require reconnection.
- Interference from other RF sources garbles reception.
- If too many phones are in use, the system is busy, and calls cannot get through.
There are three root causes of these link quality problems in wireless networks:
- Attenuation: When end points are too far away from one another they cannot hear each other. Obstacles or distance can prevent a radio signal from reaching a receiver.
- Interference: Other radios or electromagnetic emitters (such as brush motors, arc-welders, and microwaves) can generate enough noise that it becomes impossible for the receiver to distinguish between the interference and the real signal.
- Multipath: Just like sounds, when radio waves bounce off reflective surfaces, such as a metal tank, they create echoes that can act as interference and create hot and cold spots for radio reception within a space.
You can overcome attenuation and interference by turning up a transmitter's power, but this doesn't solve the multipath problem—the echoes intensify instead. Plus, because one radio's signal is another's interference, boosting the power generates more interference across the board.
Where continuous reliability is of prime importance, these problems are deadly to an automation system. Thus, despite the increasing popularity of IEEE 802.11 wireless local area network (LAN) systems and the promise of short-range Bluetooth systems, designers did not plan solutions based on these standards with the industrial environment in mind. To fulfill the promise of wireless for sensing and control systems, industrial users need a network architecture that takes the unique challenges of the industrial environment into account. But what does a real, reliable, industrial-strength wireless network really need?
Over an eighteen-month period, Ember Corp., a Boston-based start-up from the Massachusetts Institute of Technology, has investigated requirements and built pilot tests for both industrial end users and original equipment manufacturers (OEMs), and the results of this work are illuminating.
A mesh network topology is the only real solution to meet these industrial-strength needs. Without turning up transmitter power, a shorter range, multihop network makes better links and creates advantageous redundant data paths. To crack the reliability puzzle and its associated needs, the answer is a decentralized wireless network supported by all its nodes, each of which acts as a router. As a whole, the network is self-organizing, self-healing, highly scalable, and perfectly suited to connect the low performance chips that compose it.
A mesh network with redundant data paths
A star topology network with base station
Case study: water treatment plant
To validate wireless mesh networks in challenging industrial environments, Ember Corp. deployed a system in a water treatment plant. The environment was typical of such facilities, with significant wireless environment hurdles—thick reinforced concrete walls segmenting giant tanks of water, with large numbers of metal pipes running between tanks—a real challenge to wireless communication.
The goal of the installation was to connect turbidity meters in the pipe gallery back to the control panel located in the control room on the third floor of the water filtration plant. We wanted to do this with minimum installation time and maximum data-link reliability. Take a look at a geographical representation of the instrumentation topology.
There were eight instruments in the large pipe gallery along with four instruments in the small pipe gallery, and the control room was located on the third floor of an attached concrete building.
Before the installation of Ember's wireless mesh network of devices, a data collection PC in the control room communicated with process instruments over a costly, but reliable wired RS-485 serial bus. The first step in converting this system to wireless was to replace this computer's bus connection with a wireless networking card connected to its serial port.
We also replaced each process instrument's bus connections with wireless networking cards, which self-configured at power-up and began attempting to send data to the control room. After we had installed wireless cards into all 12 instruments, it was possible to analyze the RF network traffic and determine where link reliability was below standards.
These areas included spots where RF signals had to pass through reinforced concrete walls and where a single link spanned two flights of metal stairs. Improving these RF links was a simple matter of dropping down additional RF relay points. This step was made possible by the network's lack of a central wireless base station and each node's ability to cooperatively relay packets on behalf of its neighbors.
After we replaced these repeater nodes, the network was complete—all in under two hours—compared to approximately twenty hours when each instrument had to be wired back to the control panel. The project required no site survey, and because we emulated the Modbus protocol throughout the water treatment instrumentation, the software on the PC did not discern any difference between the wireless communication network and the wired serial cable network.
The wireless network exhibited less than 0.1% packet loss before we attempted to resend lost packets through the network. We accomplished this via the mesh networking algorithms the wireless network used. Neighboring nodes cooperatively relay packets over the best RF link, so the packet always finds its mark—even if the RF noise floor changes with interference or the environmental conditions.
This new networking solution brings hope for wireless in industry. This year alone will see every large industrial OEM and end user looking for wireless solutions to give voice to their products and processes. The smartest ones will be looking at embedded wireless mesh networking technologies. W
Behind the byline
Adrian Tuck is executive vice president, and Charlotte Burgess Auburn is marketing manager at Ember Corp. in Boston. (www.ember.com )
What makes an industrial wireless-device network a success?
The network cannot require sophisticated planning or site mapping to achieve reliable communications. Installers of networked devices should not have to be specialists in wireless; they are more likely to be experts in their own fields. The protocol running in the devices should do all the networking work, leaving the installation engineers to focus on what they know best.
Human intervention should not be necessary for the network to move a packet of data from one end to the other. The network should figure this out by itself. This self-configuring and self-healing aspect is vital for reliability in harsh industrial environments. All devices must be able to transmit from where they are and not have to be moved. Nobody uses wireless for fun. There has to be a pragmatic reason to switch—namely lower cost and ease of installation. If it is not easier and less expensive than copper, then the promise of wireless will never materialize. This limits the processor footprint available to such networked devices. It is important to be able to work with the much cheaper, ubiquitous 8-bit processor.
A network must be reliable. The network error rate should be below acceptable levels, as defined by the customer. It should be scaleable. Device networks of tens of thousands of end points should be possible and should not require detailed network planning. What about security? Especially when isolated wireless networks are bridged to existing networks and, ultimately, to the Internet, security is paramount. By linking remote devices to the Internet, they assume the vulnerability of any Internet-enabled structure as they come online. It is important that each device contain the capability to invoke appropriate strengths of encryption for their applications.
Range is important, but only in the context of a typical industrial network. While the span of the entire network often needs to cover several tens of thousands of square feet, the average distance between end points (nodes) is usually only a few tens or hundreds of feet. Power is not always available at the sensing or control point. Successful industrial wireless devices will often need to run on scavenged power or operate for several years using battery technology.
Successful industrial wireless networks need to coexist with the wired world. It is important that they be able to emulate the characteristics of wired networks. A true test of a successful wireless network is one where a user unplugs the wired network from a device and its controllers and plugs in a set of wireless devices. The network should be able to configure itself and pass data without needing to reengineer the devices themselves. This is the key to accelerated adoption.